Large-scale underwater fish recognition via deep adversarial learning

Zhang, Zhixue; Du, Xiujuan; Jin, Long; Wang, Shuqiao; Wang, Lijuan; Liu, Xiuxiu

doi:10.1007/s10115-021-01643-8

Cited by 21 publications

(11 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also implemented the approach of Jalal et al [ 22 ] based on the second branch of their hybrid approach, where they used YOLOv3 to detect and classify fish images. Zhang et al [ 32 ] proposed AdvFish, which addresses the noisy background problem. They fine-tuned the ResNet50 model by adding a new term in the loss function.…”

Section: Resultsmentioning

confidence: 99%

“…We tested the approach of FishResNet [ 31 ] by using the same test set, we obtained 95.62%. We used the provided code in AdvFish [ 32 ], and we trained ResNet50 by using 7-Fold cross-validation; we achieved 90.99%. In our work, without data augmentation, we obtained an AC value of 99.21% and AP value of 95.38% .…”

Section: Resultsmentioning

confidence: 99%

“…CNNs have also shown high performance for underwater vision enhancement [ 20 , 21 ], and for fish detection [ 5 , 11 , 22 , 23 ]. For fish species classification, the first works based on CNNs applied well-known pretrained networks such as AlexNet [ 24 , 25 , 26 ], VGGNet [ 27 ], GoogleNet [ 28 , 29 ] and ResNet [ 30 , 31 , 32 ]. Our approaches use ResNeXt 101 network with 101 layers [ 33 ].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Targeted Data Augmentation and Hierarchical Classification with Deep Learning for Fish Species Identification in Underwater Images

2022

View full text Add to dashboard Cite

In this paper, we address fish species identification in underwater video for marine monitoring applications such as the study of marine biodiversity. Video is the least disruptive monitoring method for fish but requires efficient techniques of image processing and analysis to overcome challenging underwater environments. We propose two Deep Convolutional Neural Network (CNN) approaches for fish species classification in unconstrained underwater environment. In the first approach, we use a traditional transfer learning framework and we investigate a new technique based on training/validation loss curves for targeted data augmentation. In the second approach, we propose a hierarchical CNN classification to classify fish first into family levels and then into species categories. To demonstrate the effectiveness of the proposed approaches, experiments are carried out on two benchmark datasets for automatic fish identification in unconstrained underwater environment. The proposed approaches yield accuracies of 99.86% and 81.53% on the Fish Recognition Ground-Truth dataset and LifeClef 2015 Fish dataset, respectively.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Targeted Data Augmentation and Hierarchical Classification with Deep Learning for Fish Species Identification in Underwater Images

2022

View full text Add to dashboard Cite

show abstract

“…Many works have also been performed to detect fish underwater but that produces further challenges based on the turbidity of the water [67,68]. Other works have been performed to classify fish including usage of the very high-resolution cameras or pictures captured by cameras dedicated to a limited number of fishes and therefore their large scale implementation is often not feasible [69,70]. These produce multiple knowledge gaps in existing studies which can be improved to build a model which can be used to detect and recognize fishes outside water at a very large scale, verified for a large number of samples, and can be implemented for both high as well as low-resolution images and this motivates to conduct this experiment.…”

Section: Motivation For the Experimentsmentioning

confidence: 99%

Fish Species Identification on Low Resolution - A Study with Enhanced Super Resolution Generative Adversarial Network (ESRGAN), YOLO and VGG-16

Adhikary¹

2022

Preprint

View full text Add to dashboard Cite

A smart detection and recognition model for the species of a fish from camera footage is of urgent requirement as fishery contributes to a large portion of the world economy and these kinds of advanced models can aid fishermen on a large scale. Such models incorporating a pick-and-place machine can be very useful to sort different fish species in a bulk without human intervention and this can greatly reduce costs for large scale fishing industries. Existing methods for detecting and recognizing fish species have many limitations such as limited scalability, detection accuracy, failing to detect multiple species, degraded performance at a lower resolution, or pinpointing the exact location of the fish. By modifying the head of a very powerful deep learning model namely VGG-16 with pre-trained weights can be used to detect both the species of the fish and find the exact location of the fish in an image by implementing a modified YOLO to incorporate the bounding box regression head. We have proposed the usage of the ESRGAN algorithm along with the proposed neural network to amplify the image resolution by a factor of 4. With this method, the overall detection accuracy of 96.5% has been obtained. The experiment has been conducted based on a total of 9460 images spread across 9 species. After further improving the model, a pick-and-place machine could be integrated for very fast sorting of the fish according to their species in different large-scale fish industries.

show abstract

“…For small and medium-scale optimization problems, EA has achieved excellent performance in various industrial application systems and can effectively handle various nonlinear, strongly coupled, mixed-variable, and other complex optimization scenarios (Xue, 2021;Xue and Jiang, 2021). However, when the scale of decision variables of the target problem exceeds a certain order of magnitude, conventional EA (Strasser et al, 2017;Ma et al, 2021b;Ma et al, 2021c;Zhang, 2022) are difficult to obtain satisfactory performances such as solution accuracy and convergence speed even with improved global optimization operator strategies due to their limited search capability (Fan et al, 2014;Ran Cheng and Yaochu Jin, 2015;Yang et al, 2017;Ma et al, 2019). Therefore, how to design efficient large-scale global optimization methods is an urgent problem to solve complex engineering system applications in big data environment.…”

Section: Introductionmentioning

confidence: 99%

Large-Scale Evolutionary Optimization Approach Based on Decision Space Decomposition

Chang²,

Yu³

2022

Front. Energy Res.

View full text Add to dashboard Cite

The identification of decision variable interactions has a crucial role in the final outcome of the algorithm in the large-scale optimization domain. It is a prerequisite for decomposition-based algorithms to achieve grouping. In this paper, we design a recognition method with higher efficiency and grouping accuracy. It is based on the decomposition strategy of min hash to solve large-scale global optimization (LSGO) problems, called MHD. Our proposed method focuses on discovering the interactions of decision variables through min hash and forming subcomponents with a principle that the interdependencies between these subcomponents are maintained at a minimal level. This is described as follows: first, the min hash performs several permutations of the vector composed of decision variables. Second, the index value of the first non-zero row of the vector after rearrangement is found to obtain the new feature vector. Third, the probability of identical data at each position is calculated based on the new feature vector to decide whether there are some certain interactions between the decision variables. The advantages of min hash are: simpler computation and greater efficiency improvement than comparison between two or two decision variables; ability to find similar decision variables very quickly; and ability to cluster decision variables in a simple way. Therefore, the efficiency as well as the reliability of MHD is guaranteed. On the accuracy aspect, the proposed algorithm performs well in various types of the large-scale global optimization benchmark test function. Finally, the experimental results analysis and summarize the performance competitiveness of our proposed MHD algorithm from several aspects when it is used within a co-evolutionary framework.

show abstract

Large-scale underwater fish recognition via deep adversarial learning

Cited by 21 publications

References 41 publications

Targeted Data Augmentation and Hierarchical Classification with Deep Learning for Fish Species Identification in Underwater Images

Targeted Data Augmentation and Hierarchical Classification with Deep Learning for Fish Species Identification in Underwater Images

Fish Species Identification on Low Resolution - A Study with Enhanced Super Resolution Generative Adversarial Network (ESRGAN), YOLO and VGG-16

Large-Scale Evolutionary Optimization Approach Based on Decision Space Decomposition

Contact Info

Product

Resources

About